Applying a common average to multiple files/recordings across multiple date and person with different bad channels

I have multiple EEG recordings with different people on different dates. Each recording has a different set of marked bad channels. Thus, if I apply a common average reference, I will not be using the same electrodes every time.

Is this actually a problem, what are the implications of this difference for future analysis that I should consider?

what do you plan to do with this data? sensor ERPs? source localization? ICA?


@agramfort There is a little bit of ERP, but that is not the main part.
The study is around neurofeedback, with an alternation of regulation (16s) and non regulation phases (8s). I don’t yet have a clear roadmap of the analysis; probably some fix length epochs and analysis to determine which area was most regulated; first in sensor space and maybe later in source space.

The lack of a clear plan is also why I’m looking now into the limitation/implications of applying a different (virtual) reference channel to all my recordings.